AI/ML Systems

AI-Powered Product Recommendation Engine

Built ML recommendation system from 0→1 serving 50,000+ users. Generated $80M incremental revenue and increased MAU by 21%.

The Problem

Wealth managers struggled to deliver personalized product recommendations at scale. Manual research took hours per client, limiting portfolio advisors to 50-100 clients each. Customers received generic suggestions disconnected from their goals, risk profiles, and financial circumstances.

The Solution

Built hybrid recommendation engine combining collaborative filtering and content-based approaches. System analyzed user behavior, portfolio composition, risk tolerance, and investment goals to generate personalized product suggestions in real-time.

Technical Implementation

Production LLM platform for AI-powered product descriptions and matching

Vector database for semantic similarity search across 500+ financial products

Collaborative filtering models trained on 50K+ user interaction patterns

A/B testing framework to measure conversion lift across recommendation strategies

Real-time serving infrastructure handling 10K+ daily recommendation requests

Results

$80M incremental new money generated in 2024

21% increase in Monthly Active B2C Users since Q1 2024 launch

27% increase in B2B2C partner engagement

Enabled advisors to scale from 50 to 200+ clients each

What Worked

Hybrid approach outperformed pure collaborative or content-based methods by 40%. Behavioral signals combined with semantic product understanding delivered higher conversion. Model complexity vs serving latency became the main trade-off at scale—chose simpler models that could serve sub-100ms for better user experience.

View all projects